|Publication number||US7848457 B2|
|Application number||US 11/177,238|
|Publication date||Dec 7, 2010|
|Filing date||Jul 7, 2005|
|Priority date||Oct 2, 2002|
|Also published as||CN1720604A, CN100536335C, EP1550149A2, EP1550149A4, US6920187, US9130793, US20040066858, US20060013337, US20110075765, WO2004032199A2, WO2004032199A3|
|Publication number||11177238, 177238, US 7848457 B2, US 7848457B2, US-B2-7848457, US7848457 B2, US7848457B2|
|Inventors||Daniel B. Penney|
|Original Assignee||Micron Technology, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (28), Classifications (21), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a continuation of U.S. patent application Ser. No. 10/264,008, filed Oct. 2, 2002 now U.S. Pat. No. 6,920,187.
The present invention relates to distribution of clock and strobe signals in semiconductor integrated circuits. In particular, the invention relates to differential receivers to receive differential clock and strobe signals.
Semiconductor integrated circuits, such as memory chips, are becoming physically larger in area, and their operational speeds are becoming faster. One challenge in such circuits is to distribute signals, such as clock and data strobe signals, over distances that are characterized as long distances, at least long as compared to the time periods available for rise times, fall times, pulse widths, etc. With such circuits, it is becoming increasingly more important to maintain correct duty cycle on certain critical signals, such as clock and strobe signals, especially at high frequencies. However, when a single pulse is propagated through many driver receiver pairs, the duty cycle (e.g., as measured by the positive or negative pulse width) may change owing in part to differences in the characteristics of, for example, the p channel and n channel transistors that make up the drivers and receivers.
One such strobe signal is a DQS (“data strobe”) signal, and the timing of the DQS signal is especially important for double data rate (DDR) DRAMs in which both the rising and falling edges of the DQS signal are used to strobe data in and out of the DRAMs. Increases or decreases in duty cycle of the strobe, for any reason, will result in data clocked in early or late. At high frequencies, data that is clocked in early, may not be accurate since the voltages on the data lines many not have had time to settle to a level that is representative of the data value.
To facilitate signals driven into heavy and multiple loads, a design may employ intermediate drivers between the signal source and the ultimate loads in a “fan out” arrangement. Intermediate drivers tend to introduce duty cycle skew as well. For example, a 3 nanosecond pulse width might become more narrow than or wider than the original 3 nanoseconds when the pulse propagates through the intermediate driver. This may be due to the threshold (or trip point) at which the intermediate driver recognizes that a received signal has switched to another state (i.e., from “1” to “0” or from “0” to “1”). This may also be due to differences in the drive of the p channel and n channel transistors that constitute the intermediate driver (also called P-N drive imbalance). These effects cause a “1” to propagate differently than a “0”.
Use of differential signal intermediate drivers can eliminate these types of duty cycle errors, but a differential receiver is required at the destination. A differential input signal, which is actually a signal and its compliment, is applied to the input of the differential receiver. A cross point of the two signals that constitute input differential signal is defined to be the voltage at which the signals cross and have the exact same voltage at the exact same time. However, the cross point may not necessarily be midway between the signal high and signal low voltage. Instead, the cross-point of the input differential signal may be high or low relative to the design sensing threshold of the differential receiver. The high or low cross point may be caused by semiconductor process variations, P-N drive differences, temperature, VCC, common mode noise coupling equally into the differential signal, etc. The function of the differential receiver is to convert an input differential signal (in which the cross point may be high or low) into a centered output signal, either a differential output signal or a single ended output signal. The output signal of the differential driver, which must finally be used as a single ended signal for CMOS logic functions, has consistent timing for the final application. Duty cycle errors can be eliminated at the receiver since one signal goes high and the other signal goes low for both logic states and the cross point for each state contains the original duty cycle information.
Differential receivers have an additional advantage. Common mode bias that couples into the differential signal from stray signals or from supply noise can cause timing error (undesired delays) in single ended drivers. However, differential receivers eliminate these timing errors. Any coupling or supply noise that is common to both differential signals can be eliminated by the differential receiver. However, a problem remains with the use of known differential receivers. Known differential receivers are based on differential transistor pairs requiring a constant current source to propagate through either transistor of the differential pair. Thus, differential receivers consume a significant current whether or not differential signals are actually needed to be processed through the receiver at the particular time. In many applications it would be impractical to use as many differential pair receivers in critical clock paths as might otherwise be desired since a high standby current would be consumed.
Another problem is that this type of differential receiver has a different propagation delay depending on the voltage of the cross point of the differential input signals (as when a common mode voltage bias is coupled into the differential input signals). A varying propagation delay is unacceptable for critical timing paths, especially at high frequencies.
There is therefore the need for a differential receiver that can generate a differential output signal from a differential input signal while preserving the timing characteristics of the differential input signal.
An integrated circuit in accordance with the invention includes a differential receiver circuit that consumes substantially no standby power, has constant propagation delay regardless of the input common mode bias, has acceptable common mode rejection and includes first and second pass circuits and buffers to receive differential input signals. The first pass circuit provides a “true” output signal when there is a difference between a “true” buffered signal and a “complimentary” buffered signal. The second pass circuit provides a “complementary” output signal when there is a difference between the “complimentary” buffered signal and the “true” buffered signal. The differential receiver circuit further includes cross coupled inverters latching the true and complimentary output signals in complementary states, and final drive buffers for amplifying the true and complimentary output signals.
The invention will be described in detail in the following description of preferred embodiments with reference to the following figures wherein:
A common mode receiver 100 according to one example of the present invention is depicted in
The first pass circuit 110 includes a p channel transistor MP1 and an n channel transistor MN1. Gates of both the transistors MP1 and MN1 are coupled to the output of the first inverter GI1 so that the signal GI1OUT is applied to the gates of the transistors MP1 and MN1. The signal GI1OUT controls the conductive state of the transistors MP1, MN1 and thus constitutes a control input of the first pass circuit 110. The sources of both the transistors MP1 and MN1 are coupled to receive the signal GI2OUT from the output of the second inverter GI2. The GI2OUT signal is coupled through one of the transistors MP1 and MN1 to the drains of the transistors MN1 and MP1 depending upon the state of the GI1OUT signal. More specifically, the GI2OUT signal is coupled through the transistors MP1 when the GI1OUT signal is low (i.e., IN is high), and is coupled through the transistors MN1 when the GI1OUT signal is high (i.e., IN is low). The drains of the transistors MP1 and MN1 are coupled to the output signal OUT and thus constitute the output of the first pass circuit 110 and the true output of the receiver 100.
Although the signal GI2OUT is passed to the output signal OUT through either of the transistors MP1 or MN1, but not both, it is not passed through the transistor MP1 until the signal GI1OUT becomes sufficiently low to turn on the transistor MP1, and the signal GI2OUT is not passed through the transistor MN1 until the signal GI1OUT becomes sufficiently high to turn on the transistor MN1. This gives the receiver 100 a desired differential effect. The differential effect derives from the fact that the transistors MP1, MN1 conduct only when the signal GI1OUT is of a different polarity than the polarity of the signal GI2OUT. The pass circuit 120 operates similarly to the pass circuit 110 to couple the signal GI1OUT through a p channel transistor MP2 to the complimentary output signal OUT* when GI2OUT is low (i.e., IN* is high), and through an n channel transistor MN2 to the complimentary output signal OUT* when GI2OUT is high (i.e., IN* is low). Again, the signal GI1OUT is not passed through the transistor MP2 until the signal GI2OUT becomes sufficiently low to turn on the transistor MP2, and the signal GI1OUT is not passed through the transistor MN2 until the signal GI2OUT becomes sufficiently high to turn on the transistor MN2. This gives the receiver 100 the desired differential effect as well as differential output signals. The differential effect derives from the fact that the transistors MP1, MN1 conduct only when the signal GI1OUT is of a different polarity than the polarity of the signal GI2OUT.
Persons of ordinary skill in the art will appreciate in light of these teachings that the first and second inverters GI1 and GI2 may advantageously be non inverting buffers. In such a configuration, signals from the buffers should be cross connected to replace the signals GI1OUT, GI2OUT in order to maintain the correct signal polarity into the pass circuits 110, 120. Furthermore, the pass circuit 120 can be omitted if a differential output signal is not desired.
The receiver 100 does not include any constant current source as may be found in a conventional differential receiver. Accordingly, the receiver 100 consumes substantially no steady state power, although it does consume transient switching power. Furthermore, the differential effect discussed above provides the receiver 100 with the ability to reject common mode bias and noise in much the same way that a conventional differential receiver is able to reject common mode bias and noise. However, unlike a conventional differential receiver, the receiver 100 affects a substantially constant time delay of the input signals that is independent of the common mode bias or noise due in part to the differential effect discussed above and due in part to the fact that the signal passing through the pass circuit, either the pass circuit 110 or 120 or both, is eventually compared to a fixed digital threshold midway between the signal maximum and minimum. The propagation delays through conventional differential receivers are known to be dependent on the common mode bias or noise.
Another embodiment of a receiver 200 according to the present invention is depicted in
The transistors in the first and second inverters GI1, GI2 and in the first and second pass circuits 110, 120 are designed to over drive the transistors in the third and fourth inverters 130, 140. Since the third and fourth inverters 130, 140 are cross coupled to form a latch having two stable states, the first and second inverters GI1, GI2 and the first and second pass circuits 110, 120 are sized to provide drive currents sufficient to cause the latch to change states if required. The drive capacity of a transistor may be characterized by the channel width to length ratio. An example of length to width ratios for the several components of the invention examples discussed herein is provided below as guidance for designing receivers according to the invention. However, it is to be understood that the drive of the several components are only representative of the invention examples, and modification may be made in light of these teachings In the example used in the simulations discussed below, each of the p channel transistors MP1, MP2 and each of the n channel transistors MN1, MN2 has a width to length ratio of 15. Each of the first and second inverters GI1, GI2 includes a p channel transistor and an n channel transistor. Each of the p channel transistors of the first and second inverters GI1, GI2 has a width to length ratio of 50, and each of the n channel transistors of the first and second inverters GI1, GI2 has a width to length ratio of 20. These preferred width to length ratios are a function of the specific semiconductor processes and design properties of the transistors involved. With this example, each of the p channel transistors MP3, MP4 has a width to length ratio of 8, and each of the n channel transistors MN3, MN4 has a width to length ratio of 6. Alternatively, the p channel transistors MP3, MP4 might have a width to length ratio of 6, and the n channel transistors MN3, MN4 might each have a width to length ratio of 4. With these parameters, the third and fourth inverters GI3, GI4 can be over powered by the drive capacity of the first and second inverters GI1, GI2 and the first and second pass circuits 110, 120. The third and fourth inverters GI3, GI4 provide the sharply latched critical load voltages that are to be further buffered as discussed below.
Another embodiment of a complete differential receiver 300 according to the invention is shown in
Another embodiment of an integrated circuit 360 on a semiconductor chip according to the invention is shown in
As discussed above, the width to length ratios as discussed herein, and the properties of the parasitics Z1 and Z2, relate only to limited invention examples that are used in the simulation discussed below, but that modifications in light of these teaching are within the spirit and scope of the invention. In keeping with the scale of the width to length ratios discussed above, each of the p channel transistors of the seventh and eighth inverters GI7, GI8 preferably has a width to length ratio of 80, and each of the n channel transistors of the seventh and eighth inverters GI7, GI8 preferably has a width to length ratio of 40. Each of the p channel transistors of the ninth and tenth inverters GI9, GI10 preferably has a width to length ratio of 250, and each of the n channel transistors of the ninth and tenth inverters GI9, GI10 preferably has a width to length ratio of 125. The drive chain 320 provides adequate drive capacity to launch a differential signal on parasitics Z1 and Z2 and from there to loads 340 beyond (e.g., one or more complete differential receivers 300). Although, the parasitic Z1 and parasitic Z2 are complexes of distributed resistances and capacitances, in aggregate, the distributed resistance of each parasitic is fairly represented by a series resistance of about 629 ohms, and the distributed capacitance of each parasitic is fairly represented by a shunt capacitance (to ground) of about 0.8926 pico Farads; however, the specific parameters depend on the specific design being considered. Each parasitic functions as a low pass filter.
In the integrated circuit 360, differential outputs of the complete differential receiver 300 are applied to the loads 340 represented by eleventh and twelfth inverter loads GI11 and GI12. Each of the eleventh and twelfth inverter loads GI11 and GI12 actually includes 5 individual inverters in a fan out arrangement, and each of these 10 inverters includes a p channel and an n channel transistor in a known inverter arrangement. These 10 inverters represent the load (e.g., the inputs of several CMOS circuits) applied to the output from the complete differential receiver 300, but the particular arrangement of this load may vary from one integrated circuit design to another. In keeping with the scale of the width to length ratios discussed above, each of the p channel transistors of the 10 inverters in the eleventh and twelfth inverter loads GI11, GI12 preferably has a width to length ratio of 40, and each of the n channel transistors of the 10 inverters in the eleventh and twelfth inverter loads GI11, GI12 preferably has a width to length ratio of 20.
This simple circuit gives reasonable performance when compared to using a differential pair based amplifier that has a constant current source and varying propagation delay.
The differential signal generator 410 may include just a buffer to receive a single ended or differential clock or strobe from outside the integrated circuit 400 and convert the clock or strobe to a differential signal before being amplified in drive chain 420. Alternatively, the differential signal generator may include logic circuits that generate the differential signal from a logical combination of signals received from outside of or inside of the integrated circuit 400.
The at least one more circuit component 460 that are not part of the integrated circuit 400 might be a CPU chip where the integrated circuit 400 is a memory device. Alternatively, the at least one more circuit component 460 that are not part of the integrated circuit 400 might be a memory device where the integrated circuit 400 is a CPU chip. In either event, system 500 includes two or more separate integrated circuit chips. These separate integrated circuits might be mounted on the same or a different printed wiring assembly before being installed into a mother board of an apparatus, or these separate integrated circuits might be mounted in any way in the same or a different apparatus and connected together by cables as in the case of a desktop computer being cable connected to a printer apparatus.
In a memory device or any other integrated circuit, or in a system on which such an integrated circuit is mounted on separate printed wiring assemblies or in separate apparatuses, the advantages over the prior art of low standby power, unvarying pulse duty cycle and repeatable propagation delays are bestowed on the integrated circuit 400 and the system 500.
Having described preferred embodiments of a novel constant delay zero standby differential logic receiver (which are intended to be illustrative and not limiting) with the details and particularity required by the patent laws, it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings. It is therefore to be understood that changes may be made in the particular embodiments of the invention disclosed which are within the scope and spirit of the invention as defined by the appended claims. Therefore, what is claimed and desired protected by Letters Patent is set forth in the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4754173||Oct 30, 1987||Jun 28, 1988||Digital Equipment Corporation||Emitter coupled logic latch with boolean logic input gating network|
|US5216298||Jan 10, 1992||Jun 1, 1993||Mitsubishi Denki Kabushiki Kaisha||ECL input buffer for BiCMOS|
|US5319259||Dec 22, 1992||Jun 7, 1994||National Semiconductor Corp.||Low voltage input and output circuits with overvoltage protection|
|US5369316||Nov 22, 1993||Nov 29, 1994||United Microelectronics Corporation||Advanced output buffer with reduced voltage swing at output terminal|
|US5386207 *||Jun 23, 1992||Jan 31, 1995||Winbond Electronics North America Corporation||Comparator with application in data communication|
|US5436934 *||Aug 13, 1992||Jul 25, 1995||3 Com Corporation||Differential high frequency level detector and data restoration circuit including squelch offset and slicing offset|
|US5486777||Sep 7, 1994||Jan 23, 1996||National Semiconductor Corporation||Low power differential receiver input circuit|
|US5549476||Mar 27, 1995||Aug 27, 1996||Stern; Sylvan S.||Method for making dental restorations and the dental restoration made thereby|
|US5654933||Jan 21, 1997||Aug 5, 1997||Micron Technology, Inc.||Equilibrated sam read transfer circuit|
|US5657266||Jun 30, 1995||Aug 12, 1997||Micron Technology, Inc.||Single ended transfer circuit|
|US5687122||Apr 29, 1996||Nov 11, 1997||Micron Technology, Inc.||Data output buffer|
|US5768178||Apr 1, 1997||Jun 16, 1998||Micron Technology, Inc.||Data transfer circuit in a memory device|
|US5825207||Jul 19, 1996||Oct 20, 1998||Kabushiki Kaisha Toshiba||Output buffer circuit|
|US5896047||Feb 5, 1997||Apr 20, 1999||Xilinx, Inc.||Balanced truth-and-complement circuit|
|US6016282||May 28, 1998||Jan 18, 2000||Micron Technology, Inc.||Clock vernier adjustment|
|US6028448||Mar 27, 1998||Feb 22, 2000||Cypress Semiconductor Corp.||Circuitry architecture and method for improving output tri-state time|
|US6049229||Jul 6, 1999||Apr 11, 2000||Pericom Semiconductor Corp.||Self-biasing CMOS PECL receiver with wide common-mode range and multi-level-transmit to binary decoder|
|US6115318||Dec 3, 1996||Sep 5, 2000||Micron Technology, Inc.||Clock vernier adjustment|
|US6351421||Feb 22, 2000||Feb 26, 2002||Micron Technology, Inc.||Data output buffer|
|US6496070||Jun 4, 2001||Dec 17, 2002||Fujitsu Quantum Devices Limited||Buffer circuit comprising load, follower transistor and current source connected in series|
|US6522160||Jun 13, 2001||Feb 18, 2003||Micron Technology, Inc.||Input buffer with automatic switching point adjustment circuitry, and synchronous DRAM device including same|
|US6552596||Aug 10, 2001||Apr 22, 2003||Micron Technology, Inc.||Current saving mode for input buffers|
|US6650157||Jan 11, 2002||Nov 18, 2003||Sun Microsystems, Inc.||Using a push/pull buffer to improve delay locked loop performance|
|US6714051||Dec 4, 2002||Mar 30, 2004||Kabushiki Kaisha Toshiba||Logic circuitry-implemented bus buffer|
|US6766155||Jan 24, 2002||Jul 20, 2004||Agilent Technologies, Inc.||Fixed termination scheme for differential receiver that compensates for process, voltage, and temperature variations|
|US6784700||Sep 3, 2002||Aug 31, 2004||Cypress Semiconductor Corporation||Input buffer circuit|
|US20020012413||Jul 10, 2001||Jan 31, 2002||Den Besten Gerrit Willem||Circuit for generating an inverse signal of a digital signal with a minimal delay difference between the inverse signal and the digital signal|
|US20030132790 *||Jan 11, 2002||Jul 17, 2003||Brian Amick||Using a push/pull buffer to improve delay locked loop performance|
|U.S. Classification||375/316, 375/283|
|International Classification||H03K5/156, H03K5/003, H04L25/10, H03K9/00, H04L25/02, H04L27/00, H03K19/003|
|Cooperative Classification||H04L25/0292, H03K19/00323, H03K5/1565, H04L25/0274, H04L25/10, H03K5/003|
|European Classification||H04L25/10, H03K5/156D, H03K5/003, H03K19/003D, H04L25/02K9, H04L25/02K3A|
|Feb 22, 2011||CC||Certificate of correction|
|May 7, 2014||FPAY||Fee payment|
Year of fee payment: 4